Testing and Understanding Second-Order Statistics of Spike Patterns Using Spike Shuffling Methods

摘要

We introduce a framework of spike shuffling methods to test the significance and understand the biological meanings of the second-order statistics of spike patterns recorded in experiments or simulations. In this framework, each method is to evidently alter a specific pattern statistics, leaving the other statistics unchanged. We then use this method to understand the contribution of different second-order statistics to the variance of synaptic changes induced by the spike patterns self-organized by an integrate-and-fire (LIF) neuronal network under STDP and synaptic homeostasis. We find that burstiness/regularity and heterogeneity of cross-correlations are important to determine the variance of synaptic changes under asynchronous states, while heterogeneity of cross-correlations is the main factor to cause the variance of synaptic changes when the network moves into strong synchronous states.